Experimental Design Single-subject

Atopic eczema in 32 children treated by massage with and without essential oils (Anderson et al., 2000) in a single case experimental design across subjects showed that this complementary therapy provided no statistically signi cant differences between the two groups after 8 weeks of treatment. This indicated that massage and thereby regular parental contact and attention showed positive... 

Sheiner and Beal (1983) presented the first study on the role of experimental design in one of their seminal papers on nonlinear mixed effects models. They showed that increasing the number of subjects improves parameter estimation accuracy, but that increasing the number of samples per subject subjects does not improve estimation to the same degree when the data were simulated from a 1-compartment model. Hence, it is better to have sparse data from more subjects than intensive pharmacokinetic data with fewer subjects. They also showed that relatively accurate and precise parameter estimates (except for residual variance) can be obtained using FO-approximation with as few as 50 subjects having a single sample collected per subject. Keep in mind, however, this was a very simple pharmacokinetic model with only two estimable parameters. 

We close this chapter with a brief introduction to the implications of work in dynamical systems theory for experimental design and analysis. This section is meant to portray a systems perspective that may be a fruitful worldview fi om which to approach research. The multivariate, replicated, repeated-measures, single-subject design can be used to provide data for examination within this dynamical systems perspective. 

Nesselroade, J. R., Jones, C. J. (1991). Multi-modal selection effects in the study of adult development A perspective on multivariate, replicated, single-subject, repeated measures designs. Experimental Aging Research, 17, 21-27. 

This is an extension of the crossover design that applies in situations where more than two treatment conditions are being evaluated. Each subject receives each treatment successively but in a different sequence. This design can be used in the case of single-dose experimental trials, such as studies of cognitive function, psychomotor performance, psyehophysiological responses, etc. in healthy volunteers (Chapter 3) but also in therapeutic short-term trials of hypnotics or anxiolytics. 

Since these randomized blocks are applied to single out inequality effects of a research subject from factor effects, the variance of analysis confidence is increased as experimental error is diminished. The block denotes the part of design points where experimental error is lower than in the experiment as a whole. 

Using the so-called fixed-ratio or ray design, the mixture of interest is analyzed at a constant concentration ratio while the total concentration of the mixture is systematically varied. Hence, a concentration-response curve (a ray in the mixture response surface see Section 4.5.2) of the mixture is recorded, which can then be analyzed just as the concentration-response curve of a single chemical. A comparison with the concentration-response curves of the individual components allows a comparison with both CA and IA predictions. For this purpose k(n + 1) test groups are needed, where n = number of mixture components and k = number of concentrations per concentration-response curve. For the fixed-ratio design the use of D-optimal designs has been proposed as an efficient approach (Coffey et al. 2005). For a fixed sample size, the D-optimal design provides a criterion to select the experimental concentrations and number of subjects per concentration level that... 

---